In his keynote speech at the HiPEAC conference, Ivo Bolsens, SVP Adaptive and Embedded Computing group, explained how scalable architectures can overcome the differences in cloud and edge computing
As an example of the challenges he explains how arithmetic can render different results in a edge compute environment, compared to a cloud environment. In an ideal world you want to have the same outcome for both worlds, but the reality can be different. For critical applications the requirements are high, you want to get similar compute results, you want low latency and low energy – requirements that are only possible when your hardware is flexible and adaptive.
Adaptive is the keyword in AI architectures from edge to cloud.
He also addresses the energy consumption in training data. With energy resources under pressure the requirements for training data are getting more and more questions? How much nodes can we allow to run in our data centres?
In the preparation of the HiPEAC conference, eeNews Europe interviewed Ivo on these developments and in the sideline we discussed the new announced MI300 chip. You can watch back the interview here: